The application of traction control systems (TCS) for electric vehicles (EV) has great potential due to easy implementation of torque control with direct-drive motors. However, the control system usually requires road-tire friction and slip-ratio values, which must be estimated. While it is not possible to obtain the first one directly, the estimation of latter value requires accurate measurements of chassis and wheel velocity. In addition, existing TCS structures are often designed without considering the robustness and energy efficiency of torque control. In this work, both problems are addressed with a smart TCS design having an integrated acoustic road-type estimation (ARTE) unit. This unit enables the road-type recognition and this information is used to retrieve the correct look-up table between friction coefficient and slip-ratio. The estimation of the friction coefficient helps the system to update the necessary input torque. The ARTE unit utilizes machine learning, mapping the acoustic feature inputs to road-type as output. In this study, three existing TCS for EVs are examined with and without the integrated ARTE unit. The results show significant performance improvement with ARTE, reducing the slip ratio by 75% while saving energy via reduction of applied torque and increasing the robustness of the TCS.